Analysis
Call it the forward-deployed-engineer problem. Crunchbase News -- the editorial arm of the deal-data platform Pulse has covered before -- laid out the case this week that a wave of enterprise "AI platform" companies aren't actually selling autonomous software -- they're sending engineers into customer sites to hand-build custom workflows, then booking the resulting fees as software annual recurring revenue. Airrived co-founder and CEO Ashu Gurtu, who's making the argument publicly, calls it "people-market fit" rather than product-market fit: "This is not a flywheel. This is a payroll with a pitch deck."
We think the numbers back up more concern than comfort. FDE job postings are up roughly 700% to 1,000% year over year by various counts -- Indeed listings for the role alone went from 643 in April 2025 to 5,330 in April 2026, a roughly 730% jump. Two examples show how much is riding on the model:
- Anthropic -- posting forward-deployed-engineer roles at $280,000-$320,000 base pay, with total compensation reportedly reaching $400,000 at the high end for a function that, by definition, doesn't scale the way software does.
- OpenAI's DeployCo -- a joint venture that has reportedly drawn more than $4 billion in investment at a $10 billion valuation from backers including TPG, Bain Capital, Advent, Brookfield and Goanna, which guarantees those private-equity backers a 17.5% annual return over five years.
“The test Gurtu proposes is simple and, we think, correct: software ARR should get cheaper to deliver as it scales; human-powered ARR doesn't.”
The test Gurtu proposes is simple and, we think, correct: software ARR should get cheaper to deliver as it scales; human-powered ARR doesn't. Every dollar of FDE-driven revenue carries a real, ongoing cost that a pure software dollar doesn't, and that gap shows up eventually in gross margin, not in the growth-rate slide.
Room for disagreement: FDEs may simply be the on-ramp this generation of enterprise AI products needs before self-serve deployment matures -- plenty of now-obviously-great software companies spent years with services-heavy delivery models before product usage caught up to the pitch. A services-heavy phase isn't automatically a bubble; it's a bubble only if the market keeps paying software multiples for it past the point where margins should have started improving.
What to watch: whether companies leaning on FDEs start disclosing services revenue separately from product ARR, and whether any of the big-name AI platforms show delivery costs falling per customer as their install base grows. Until then, the multiple is doing the work the margin hasn't earned yet.